Oligotyping Reveals Community Level Habitat Selection Within the Genus Vibrio

Oligotyping reveals community level habitat selection within the genus Vibrio Victor Schmidt, Julie Reveillaud, Erik Zettler, Tracy Mincer, Leslie Murphy, Linda Amaral-Zettler

To cite this version:

Victor Schmidt, Julie Reveillaud, Erik Zettler, Tracy Mincer, Leslie Murphy, et al.. Oligotyping reveals community level habitat selection within the genus Vibrio. Frontiers in Microbiology, Frontiers Media, 2014, 5 (563), pp.1-14. 10.3389/fmicb.2014.00563. hal-01104930

HAL Id: hal-01104930 https://hal.univ-brest.fr/hal-01104930 Submitted on 19 Jan 2015

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. ORIGINAL RESEARCH ARTICLE published: 13 November 2014 MICROBIOLOGY doi: 10.3389/fmicb.2014.00563 Oligotyping reveals community level habitat selection within the genus Vibrio

Victor T. Schmidt 1,2, Julie Reveillaud 1†, Erik Zettler 3, Tracy J. Mincer 4, Leslie Murphy 1 and Linda A. Amaral-Zettler 1,5*

1 Marine Biological Laboratory, Josephine Bay Paul Center for Comparative Molecular Biology and Evolution, Woods Hole, MA, USA 2 Department of Ecology and Evolutionary Biology, Brown University, Providence, RI, USA 3 Sea Education Association, Woods Hole, MA, USA 4 Department of Marine Chemistry and Geochemistry, Woods Hole Oceanographic Institution, Woods Hole, MA, USA 5 Department of Earth, Environmental and Planetary Sciences, Brown University, Providence, RI, USA

Edited by: The genus Vibrio is a metabolically diverse group of facultative anaerobic bacteria, Rachel Susan Poretsky, University of common in aquatic environments and marine hosts. The genus contains several species Illinois at Chicago, USA of importance to human health and aquaculture, including the causative agents of human Reviewed by: cholera and fish vibriosis. Vibrios display a wide variety of known life histories, from Patrick K. H. Lee, City University of Hong Kong, China opportunistic pathogens to long-standing symbionts with individual host species. Studying Migun Shakya, Dartmouth College, Vibrio ecology has been challenging as individual species often display a wide range of USA habitat preferences, and groups of vibrios can act as socially cohesive groups. Although *Correspondence: strong associations with salinity, temperature and other environmental variables have been Linda A. Amaral-Zettler, Marine established, the degree of habitat or host specificity at both the individual and community Biological Laboratory, Josephine Bay Paul Center for Comparative levels is unknown. Here we use oligotyping analyses in combination with a large collection Molecular Biology and Evolution, of existing Vibrio 16S ribosomal RNA (rRNA) gene sequence data to reveal patterns 7 MBL St., Woods Hole, MA 02543, of Vibrio ecology across a wide range of environmental, host, and abiotic substrate USA associated habitats. Our data show that individual taxa often display a wide range of e-mail: [email protected] habitat preferences yet tend to be highly abundant in either substrate-associated or †Present address: Julie Reveillaud, UMR free-living environments. Our analyses show that Vibrio communities share considerable 6197-Laboratoire de Microbiologie overlap between two distinct hosts (i.e., sponge and fish), yet are distinct from the abiotic des Environnements Extrêmes plastic substrates. Lastly, evidence for habitat specificity at the community level exists (LM2E), Institut Universitaire in some habitats, despite considerable stochasticity in others. In addition to providing Européen de la Mer (IUEM), Université de Bretagne Occidentale, insights into Vibrio ecology across a broad range of habitats, our study shows the utility Plouzané, France; of oligotyping as a facile, high-throughput and unbiased method for large-scale analyses CNRS, UMR6197-Laboratoire de of publically available sequence data repositories and suggests its wide application could Microbiologie des Environnements Extrêmes (LM2E), Institut greatly extend the range of possibilities to explore microbial ecology. Universitaire Européen de la Mer Keywords: oligotyping, Vibrio ecology, host-microbe interactions, illumina sequencing, 16S rRNA analysis, (IUEM), Place Nicolas Copernic, plastisphere, aquaculture pathogens, meta-analysis Plouzané, France; Ifremer, Centre de Brest, REM/EEP/LM2E, ZI de la Pointe du Diable, CS 10070, Plouzané, France

INTRODUCTION over 570 publicly available annotated genomes and over 64,000 Vibrio is a ubiquitous, speciose and commercially important 16S rRNA gene sequences annotated as vibrios in GenBank as bacterial genus with both host associated and free-living represen- of March 2014. Vibrio therefore represents an ideal candidate tatives. Several species within the genus are pathogenic to humans for applying new analytical approaches using pre-existing data to and animals. Vibrio cholerae has caused six historic and one ongo- gain further insights into the ecology of the genus. ing cholera pandemic, and countless epidemics (Mutreja et al., Making sense of Vibrio ecology has been a challenge, owing in 2011) including a recent outbreak in Haiti that killed more than part to its complex life history, its capacity to partition resources, 8000 people (Chin et al., 2011). Vibrio pathogens are also impor- and a strong propensity for lateral gene transfer between closely tant to the aquaculture industry, where they inflict costly losses related species (Hunt et al., 2008; Cordero et al., 2012). The com- on farmed fish, mollusks and shrimp (Austin and Austin, 2007), plexityofthegenusiswellillustratedbythediversityofitslife limiting the development of an industry poised to help bridge histories. On one hand, Vibrio has an average of 11 rRNA gene global food gaps and preserve wild fisheries (FAO, 2012). Due to copies, allowing for rapid growth rates under good conditions their importance to human and animal welfare, and the ease with (Heidelberg et al., 2000), suggestive of r-selected taxa, which can which they are cultured, vibrios are relatively well studied, with rapidly multiply given favorable conditions (Andrews and Harris,

www.frontiersin.org November 2014 | Volume 5 | Article 563 | 1 Schmidt et al. Vibrio community structure across habitats

1986). Conversely, bioluminescent vibrios have formed symbiotic involved culturing isolates from target habitats and sequencing relationships with squid and anglerfish over evolutionary time multiple loci in order to gain sufficient taxonomic resolution scales (Ruby and Nealson, 1976), suggestive of a more stable K- within a sample, requiring the use of Vibrio-specific primers selected strategy. Some flexibility between r vs. K strategies may (Preheim et al., 2011a; Szabo et al., 2013), and making-large scale, even exist within fine scale taxonomic categories, as environmen- non-targeted, multi-habitat analyses challenging and costly. More tal conditions such as pH, concentrations of bile, bicarbonate recently, oligotyping rRNA gene amplicon sequences affords and nutrients may trigger rapid growth within a host in a for- extremely high resolution analysis of community structure by mally dormant environmental bacterium (Skorupski and Taylor, selecting a subset of highly informative nucleotide sites within 2013). To further complicate the ecology of individual Vibrio single loci of 16S rRNA gene hypervariable regions alone (Eren species, recent experiments indicate that disparate species can et al., 2013a, 2014; Reveillaud et al., 2014). At the same time, form socially cohesive groups, taking advantage of their propen- repositories of 16S rRNA gene sequences have grown in size and sity for exchanging genetic elements to confer greater antibiotic scope. The Visualization and Analysis of Microbial Population resistance among closely related strains, and to likely regulate Structures (VAMPS) database is one such repository that contains virulence (Cordero et al., 2012). over 1000 datasets representing hundred of millions of publically Vibrios also seem to be highly variable in habitat preference. available 16S rRNA gene sequences (Huse et al., 2014). Traditionally Vibrio life history has been studied in association The aim of the present study was to use oligotyping to explore with multicellular marine hosts, including fish, mollusks, and a the distribution of Vibrio communities in a range of substrate- wide range of zooplankton (Liston, 1956; Aiso et al., 1968), yet associated (both biotic and abiotic) and free-living aquatic envi- they can also exist in the ambient aquatic environment, associ- ronments. We used this method to test the hypothesis that distinct ated with plastic particles (Zettler et al., 2013), or phytoplankton Vibrio communities occur in different habitats, and are char- blooms (Gilbert et al., 2012). Whether individual Vibrio species, acterized by clear distinctions between host habitats and their or communities of vibrios, are specific to particular habitats is an surrounding water. open question, and distinguishing specialized associations from opportunistic colonization is challenging (Takemura et al., 2014). MATERIALS AND METHODS Host specificity has been observed in other bacterial genera, SEQUENCE COLLECTION including Blautia (Eren et al., 2014)andNitrospira (Reveillaud The VAMPS database houses 16S rRNA gene amplicon sequence et al., 2014), but because Vibrio is abundant in both host and envi- data projects from a wide variety of environmental and host- ronmental habitats, distinguishing established host associations associated habitats. We identified seven existing projects to tar- from incidental or ephemeral colonization from surrounding get for analyses of Vibrio diversity, representing free-living and habitats is difficult. host (abiotic and biotic) substrate associated habitats (Table 1). Because vibrios are diverse in their habitat preferences and We chose projects with the occurrence of at least three sam- potentially act as socially cohesive units, large-scale analysis of ples with >300 sequences identified as Vibrio by the Global Vibrio community structure across habitats may provide impor- Alignment for Sequence Taxonomy (GAST) pipeline (Huse et al., tant insights into its ecology. Analyses of this type have historically 2008) in the VAMPS database. In rare cases, a sample was

Table 1 | Overview of projects used from the VAMPS database with their original citation.

VAMPS project Habitat Sample Mean Vibrio Geographic location Salinity Citation or SRA BioProject number relative abundance accession number

ICM_PML_Bv6 Seawater 3 0.31 (SE 0.13) English Channel Marine Gilbert et al., 2012 LAZ_MHB_Bv6 Seawater 14 0.0017 (SE 0.00021) Northwestern Atlantic Marine SRP049014 LAZ_NMS_Bv6 Saltmarsh 11 0.173 (SE 0.019) New England, USA Mixed SRP059013 SLM_NIH_Bv6 PAH spiked sand 11 0.012 (SE 0.002) Gulf of Mexico Mixed Kappell et al., 2014 LAZ_SEA_Bv6 Seawater - associated 32 0.0036 (SE 0.00058) Northwestern Atlantic Marine SRP026054 with plastic LAZ_SEA_Bv6 Plastic-associated 27 0.0032 (SE 0.0005) Northwestern Atlantic Marine SRP026054 JCR_SPO_Bv6 Seawater - associated 11 0.055 (SE 0.023) Northeastern Atlantic Marine Reveillaud et al., 2014 with sponge JCR_SPO_Bv6 Sponge-associated 49 0.09 (SE 0.014) Northeastern Atlantic Marine Reveillaud et al., 2014 VTS_MIC_Bv6 Aquarium water - 31 0.016 (SE 0.0037) MBL, Woods Hole, USA Mixed SRP047374 (but see associated with ﬁsh Supplementary Data Sheet 1) VTS_MIC_Bv6 Fish-associated 20 0.3 (SE 0.039) MBL, Woods Hole, USA Mixed SRP047374 (but see Supplementary Data Sheet 1)

The mean Vibrio relative abundance across all samples in a given project is shown with standard error. For sequences ﬁrst published by this study the accession numbers for that project’s NCBI Sequence Read Archive (SRA) BioProject is given.

Frontiers in Microbiology | Systems Microbiology November 2014 | Volume 5 | Article 563 | 2 Schmidt et al. Vibrio community structure across habitats

included that had less than 300 sequences in order to increase We then extracted microbial gDNA from the homogenate using the number of samples for habitats with low sample num- a modified Gentra Puregene Yeast/Bac (Qiagen, Valencia, CA) bers, or when water associated with a specific substrate was of extraction protocol (Supplementary Data Sheet 1). We collected interest but possessed low Vibrio sequence representation. All microbial communities from tank water using sterile 1 L PET bot- projects were sequenced at the Marine Biological Laboratory tles, and extracted gDNA according to protocols outlined above (MBL) Keck sequencing facility on an Illumina HiSeq 1000, and for LAZ_MHB_Bv6 samples. employed identical protocols in the generation and sequencing of 16S rRNA gene sequences, including the same primer OLIGOTYPE GENERATION AND ANALYSIS cocktails to target the V6 hypervariable region, as described else- Oligotyping is a supervised method that allows the identification where (Eren et al., 2013b). Each project also followed the same of closely related but distinct bacterial taxa in high-throughput standard MBL sequence analysis pipeline, where only perfectly sequencing datasets of marker genes. This novel bioinformatics overlapping paired-ends reads with zero mismatches passed qual- approach is capable of uncovering ecological patterns of micro- ity filters (Eren et al., 2013b), and taxonomic assignment was bial communities at finer scales than previously possible with done using the GAST pipeline (Huse et al., 2008). Both qual- de novo approaches (Eren et al., 2013a). Oligotying exploits the ity filtering and taxonomic assignments had already been made fact that some positions within a DNA marker sequence are for all sequences across all projects as part of standard VAMPS more ecologically informative than others. The method identifies protocols, and we were therefore able to directly download highly variable locations using Shannon entropy (that is, “entropy sequences using VAMPS’s “data export -> TaxBySeq” feature, components”), and uses only these positions to discriminate eco- using the query “Bacteria; Proteobacteria; Gammaproteobacteria; logical units, so called oligotypes. This process reduces the impact Vibrionales; Vibrionaceae; Vibrio.” of noise caused by sequencing error by relying on only a small Although sequence generation was identical for all projects, number of nucleotide positions, discarding the redundant parts sample collection varied. Saltmarsh water sample collec- of reads for the identification of oligotypes. The open-source tion (LAZ_NMS_Bv6, this study) employed automated collec- pipeline for oligotyping is available from http://oligotyping.org. tion via the Phytoplankton Sampler (PPS) (McLane Research This method has been used previously to identify Gardnerella dis- Laboratories, Inc., East Falmouth, MA) that filters 500 mL of tributions in vaginal samples (Eren et al., 2011), Nitrospira speci- water through a 0.65 µm flat filter (EMD Millipore Durapore ficity in sponges (Reveillaud et al., 2014), and Blautia specificity PVDP hydrophilic membrane filters) (Billerica, MA) twice a day, in animal hosts (Eren et al., 2014). and stores the filter in RNAlater (Qiagen, Valencia, CA) buffer. In order to get the best possible insights into Vibrio oligo- As part of a broader project to understand microbial popu- type distributions across habitat types, we grouped samples into lations in coastal environments, we deployed PPS samplers in three broad analysis groupings; substrate (host) associated habi- tidal creeks (Mill and Salt Ponds, Nauset Marsh System, MA) tats only, substrate habitats along with their surrounding water that receive daily tidal fluxes from the Atlantic Ocean off Cape samples, and environmental and substrate associated habitats Cod, MA. DNA extraction and purification of filters used a (Table 2). First, we analyzed only substrate-associated samples modified salt precipitation method with bead-beating (Gentra (fish, sponge and our abiotic substrate—plastic marine debris). Puregene, Qiagen, Valencia, CA). The Rhode Island Department We subsampled these datasets to the median Vibrio sequence of Environmental Management (RIDEM) collected seawater sam- count of 30,000 prior to analysis in order to minimize the range ples from the northeastern reach of Narragansett Bay called in initial sequencing depth between samples, which can otherwise Mount Hope Bay, MA (LAZ_MHB_Bv6) as part of their monthly reduce the entropy value of discriminating points in datasets with water quality survey for shellfish safety. Samples collected manu- much lower sequence counts. We then processed them through ally from surface waters in sterile 1 L polyethylene terephthalate the oligotyping pipeline, as described in Eren et al. (2013a). (PET) bottles at 17 stations throughout the 36 km2 bay were sub- To establish entropy components that fully decomposed our sequently filtered through 0.22 µm polyethersulphone membrane sequence data, we started with the strongest two components, Sterivex filters (Millipore, Billerica, MA) followed by DNA extrac- then manually chose the next component that best removed tion as above. Collection details for other samples and metadata remaining entropy in the resulting oligotypes, and re-ran the are found in respective publications (Table 1). analysis. This iterative process yielded a final 12 entropy points Samples from fish and fish tanks (VTS_MIC_Bv6) were col- that fully decomposed our sequences. We allowed for oligotypes lected as part of an experiment to understand the role of salinity to occur in only a single sample (-s 1) but discarded them if and external microbiota on fish microbiomes (Schmidt et al., they did not represent at least 0.5% of the relative abundance of Submitted) (Supplementary Data Sheet 1). We acclimated ∼1 that sample (–a 0.5). Not all samples contained 30,000 sequences, inch Black Molly fish (Poecilia sphenops) to four salinity levels and sequencing depth ranged from 83 to 30,000 after rarefaction (salinities 0, 5, 18, and 30) over 30 days using nanopure water (Table 2). and Instant Ocean® (Blacksburg, VA) salt mix, then maintained Global alignment prior to oligotyping for short Illumina reads each fish at target salinity for 12 days. Each salinity treatment is unnecessary, as positional shifts in sequencing reads due to nat- contained four independent tanks, each with two fish. For our ural indels will produce entropy peaks at the position of insertion analyses here, we grouped salinities 0 and 5 (FreshwaterFish) or deletion (and subsequent positions), and the decomposition and salinities 18 and 30 (MarineFish). After 12 days we eutha- of the dataset based on any of these peaks will eventually result in nized fish in 1 mg/mL MS-222 and homogenized the entire fish. the same oligotypes as if they would have been previously aligned.

www.frontiersin.org November 2014 | Volume 5 | Article 563 | 3 Schmidt et al. Vibrio community structure across habitats

Table 2 | Description of samples for each analysis grouping.

Analysis grouping Analysis Number of Subsampled Oligotypes Percentage of reads Components (position grouping/ samples sequence before (after) represented by top 5% in alignment) Figure depth range quality ﬁltering (10%) oligotypes

All substrates 1 104 83–30,000 882 (74) 76 (94) 13, 15, 20, 21, 22, 23, 25, 31, 32, 45, 50, 55 Plastics and surrounding 2C 71 83–13,359 415 (71) 90 (96) 13, 15, 20, 21, 22, 23, 25, seawater 31, 32, 45, 50, 55 Sponges and surrounding 2B 58 1822–10,000 681 (45) 71 (90) 13, 15, 20, 21, 22, 23, 25, seawater 31, 32, 45, 50, 55 Fish and surrounding 2A 51 636–85,000 604 (21) 80 (97) 13, 15, 20, 21, 22, 23, 25, water 31, 32, 45, 50, 55 Mixed habitat 4 179 83–20,000 1452 (99) 65 (90) 13, 15, 20, 21, 22, 23, 25, 31, 32, 37, 45, 50, 55

Some samples were included in more than a single grouping.

Furthermore, V6 primers target a hypervariable region with few grouping. Since most oligotypes were abundant across multiple insertions or deletions, and Illumina technology does not have projects, this list collapsed into 17 unique oligotypes across all indel error issues. We therefore did not create an alignment prior three analysis runs. These 17 oligotypes were assigned identifiers to entropy analyses. Instead, we aligned sequences at their 3 end (“Oligotype1” through “Oligotype17”) that remained consistent and padded any length discrepancies with gaps at the 5 end prior across all three runs (Table 3 and Table S1). to entropy analysis, as detailed in Eren et al. (2013a).Wedonote that a single Oligotype, Oligotype 10, was not fully decomposed VISUALIZATIONS AND STATISTICAL ANALYSES OF OLIGOTYPE (Supplementary Data Sheet 1). We note that this oligotype varies DISTRIBUTIONS widely from all other Vibrio sequencesinthisstudy,andwould To v isualize Vibrio community similarity between samples and require an additional 4 entropy components to fully decompose. habitats we constructed Nonmetric Multidimensional Scaling Furthermore, it occurred in high abundance only in the Sand- (NMDS) plots as part of our oligotyping pipelines. We included PAH habitat. We make no conclusions about this oligotype across the covariance ellipsoids calculated as part of the oligotyping any habitat. pipeline on these plots to visualize the spread of a given habitat’s Next, we examined each substrate or host sample alongside community variance. Importantly, covariance ellipsoids delineate its respective water sample. For fish and plastics, water sam- the total high-dimensional space, not only the two axes shown ples were directly associated with the host, and collected at the in the NMDS plot. Statistical analyses of Vibrio oligotype dis- same time and place as host material. Sponge and corresponding tributions between and within habitat types followed a 4-step seawater were collected simultaneously, although not all sponge analysis using the ecological statistics packages PrimerE v.6 and samples have a corresponding water samples (see Reveillaud et al., the R package Vegan (Oksanen et al., 2012). First, we normal- 2014). For each analysis, we subsampled Vibrio sequences to the ized an oligotype matrix, which consisted of samples across rows median sequencing depth. The same 12 entropy components and and oligotypes down columns (Supplementary Data Sheet 2), oligotyping parameters as above fully decomposed all but one by percent per sample (i.e., to 100% total for each sample) and oligotype, and were therefore used again in this analysis. calculated pairwise Bray-Curtis similarities. Second, we assigned We then analyzed oligotype distributions across the broadest each sample to a habitat “factor” based on where it was col- range of samples and projects included in this study in a single lected (e.g., on plastics, sponges or seawater) and tested the null analysis using the same methods and 12 entropy components, hypothesis that there were no community differences between with one additional component added to fully resolve novel oligo- habitat types using Analysis of Similarity (ANOSIM) permuta- types from additional samples (13 components total). The added tion tests. This test builds a random distribution of oligotype samples included water samples from saltmarshes (Saltmarsh), abundances using 9999 permutations then assesses the likelihood seawater from a large coastal bay (Seawater), open ocean seawa- that observed oligotype distributions across aprioriassigned ter (Seawater), and sand samples from oiled beaches in the Gulf habitat factors occurred by chance. We then conducted pair- of Mexico inoculated with Polycyclic Aromatic Hydrocarbons wise ANOSIM tests to determine whether significant differences (PAHs) (Sand-PAH) (Tables 1, 2). As above, samples were sub- occurred between individual habitat factors. Third, when statis- sampled down to the median value of 20,000. An interactive tical groupings did occur (as they did in most cases), we iden- html file of the results from this oligotyping analysis grouping tified the oligotypes that contributed most to the formation of (Mixed Habitat) is included in the Supplementary Material under these groupings using Similarity Percentages (SIMPER). SIMPER “html_files/html.index” (Supplementary Data Sheet 2). decomposes average Bray-Curtis similarities between all pair- Finally, we grouped the representative sequences from the wise habitat comparisons into percentage contributions of each 10 most abundant oligotypes (30 total) from each analysis oligotype.

Frontiers in Microbiology | Systems Microbiology November 2014 | Volume 5 | Article 563 | 4 Schmidt et al. Vibrio community structure across habitats

Table 3 | Summary of the 10 most abundant oligotypes from each of three oligotyping analysis groupings.

MEGABLAST results Percentage of each isolation source category for MEGABLAST hits

Oligotype ID Number of 100% Top species level hits 1 2 3 4 5 No data Habitats with Similarity hits to nr database (number of hits to that Percentage (SIMPER) species) results > 10%

Oligotype 1 22 V. alfacsensis (6) 45.59.140.90.00.04.5 Sponge/MarineFish/ V. sinaloensis (2) MarineWater Oligotype 2 171 V. metschnikovii (3) 42.915.312.411.20.018.2 FreshwaterFish/ V. neptunius (11) FreshWater Oligotype 3 780 V. scophthalmi (4) 55.711.46.90.10.025.9 MarineFish/MarineWater/ V. ichthyoenteri (31) Seawater/Saltmarsh/ V. anguillarum (37) Plastic Oligotype 4 39 V. cholerae (9) 2.60.00.00.030.866.7 FreshwaterFish/Fresh V. vulnificus (17) Water/Sponge/MarineFish V. mimicus (5) Oligotype 5 1000* V. splendidus (52) 45.321.48.10.10.025.1 Sand-PAH/Seawater/ V. mediterranei (35) Saltmarsh/Plastic V. gigantis (39) Oligotype 6 23 V. ichthyoenteri (1) 70.88.30.00.00.020.8 Sponge/MarineFish/ V. ordalii (1) MarineWater Oligotype 7 46 V. ponticus (12) 30.48.74.30.00.056.5Seawater V. nigripulchritudo (9) Oligotype 8 221 V. cholerae (184) 1.41.40.53.611.881.4 FreshwaterFish/ FreshWater Oligotype 9 1 V. vulnificus (1) 0.00.00.00.0100 0.0 Oligotype 10 18 V. alginolyticus (1) 11.10.00.089.00.00.0 Sand-PAH Oligotype 11 1 None 0.00.00.00.00.0100 Oligotype 12 113 V. coralliilyticus (2) 27.439.80.00.00.032.7 Seawater/Plastic Oligotype 13 66 V. azureus (13) 54.56.10.011.65.522.7 Plastic V. h a r vey i (2) Oligotype 14 0 V. azureus (5) 0.00.00.00.00.0100.0 V. owensii (2) Oligotype 15 245 V. vulnificus (4) 94.52.40.40.00.03.4 V. shilonii (3) Oligotype 16 140 V. kanaloae (3) 42.114.311.42.10.030.0 V. splendidus (2) Oligotype 17 6 V. splendidus (3) 84.00.00.00.00.016.0

Overlap in the most abundant sequences between groupings reduced the total number to 17. Oligotype names were assigned arbitrarily, but are consistent across all groupings. The species assignments given by reports from 100% MEGABLAST hits, and the number of hits to each species, is shown. The proportion of MEGABLAST hits isolated from each of the ﬁve habitat categories are also shown. Categories are; 1. Marine Host, 2. Seawater, 3. Other Marine, 4. Terrestrial or Human, and 5. Freshwater (see Materials and Methods). (*) Indicates maximum requested hits.

To gain insight into the taxonomy and ecology of our olig- created a PhyML tree of existing full-length Vibrio 16S rRNA otype sequences, we isolated the representative sequence from gene sequences downloaded from type strains in the SILVA ARB the 17 most abundant oligotypes outlined above. We then used v5.1 database and then added our oligotype sequences to this MEGABLAST to query these sequences against National Center tree using the Maximum Parsimony feature in ARB across the V6 for Biotechnology Information’s (NCBI) nr database in June 2014 region only (using a V6 “ﬁlter” in ARB). (nr = non-redundant amalgamation of GenBank, RefSeq, EMBL, Our rationale for building this tree was not to reconstruct phy- DDBJ and PDB databases). We kept only 100% matches across logenetic relationships between Vibrio species but rather to make the entire 60 bp query, and extracted the “isolation source” and some inference about the habitats from which closely related “host” feature using Geneious (v. 6.1) annotation tables. We Vibrio 16S rRNA gene sequences have been isolated. To this end, also extracted the most abundant 2 or 3 taxonomies from per- we binned the isolation source and host annotations of both fect hits, not including “uncultured bacterium” (Table 3). We our oligotype MEGABLAST hits, and our ARB isolates, into ﬁve

www.frontiersin.org November 2014 | Volume 5 | Article 563 | 5 Schmidt et al. Vibrio community structure across habitats

habitat categories. These were “Marine Host,” “Seawater,” “Other onesample(>1% relative abundance) were always found across Marine,” “Terrestrial,” and “Freshwater.” Marine host included all all three substrate types, meaning abundant oligotypes were sequences isolated from the skin or innards of a host in seawa- ubiquitous across all host-associated habitats. Oligotype rich- ter (e.g., tunicates, fish, crustaceans, sponges and sea cucumbers). ness varied across host/substrate type. Of the 74 total oligotypes, “Other Marine” included sediment, biofilms and algae, or other 24.1 ± SE 2.2 were found in FreshwaterFish, 27.0 ± SE0.71 in marine plant associated sources. “Terrestrial” included any sam- MarineFish, 31 ± SE 0.9 in Sponge and only 18 ± SE1.5 for Plastic ple taken from the terrestrial environment, or from a terrestrial samples. host (e.g., humans, birds, and plants), and freshwater included Pairwise Analysis of Similarity (ANOSIM) tests showed hits isolated from a freshwater environment, or freshwater host significant groupings in oligotype communities according (e.g., freshwater shrimp). We color-coded the proportion of each to habitat, except between high salinity fish (MarineFish) category and displayed them at the terminal nodes of each oligo- and sponges, whose communities could not be significantly type sequence in our PhyML reference tree. Lastly, to visualize the distinguished. This ANOSIM result is also visible in our isolation source of the ARB isolate reference sequences, we color- NMDS analyses which shows clear overlap between both coded the nodes of the tree according to the isolation source of sponges (DeepSponge and ShallowSponge) and high salin- each reference sequence at the terminal node of that branch. We ity fish (MarineFish), yet separation from plastic and low used the online software Interactive Tree of Life (iTOL) (Letunic salinity fish (FreshwaterFish) habitats (Figure 1). An ANOSIM and Bork, 2011) to visualize the phylogenetic tree. The propor- test comparing marine biotic substrates (MarineFish and tion of NCBI hits for each oligotype that fit into each category Sponges pooled together) revealed a significant grouping that also appear in Table 3. excluded Plastic, an abiotic substrate. Similarity Percentages (SIMPER) analysis corroborated these results by illustrating “WITHIN-HABITAT VARIANCE” OF VIBRIO COMMUNITIES thestrongcontributionofOligotypes1,4,and6toboth To determine differences in habitat specificity, we calculated the MarineFish and Sponge within-habitat similarity, and dis- median Vibrio community variance across all datasets within the tinguished those habitats from FreshwaterFish and Plastics. same habitat (its “within-habitat variance”), and compared that Oligotypes 2, 4, and 8 contributed to both within habitat across habitats. This allowed us to determine if some habitats similarity, and between habitat differences for FreshwaterFish contained a specific Vibrio community, or if Vibrio communi- samples. Plastics were dominated by Oligotype 5, which also dis- ties varied widely even within the same habitat. To calculate tinguished it from other habitat types, including biotic substrates the median variance of all datasets in a habitat, we normalized (Tables 4A,B). our oligotype abundance matrixes by the maximum value of each oligotype, then calculated Bray-Curtis community similarity OLIGOTYPE DISTRIBUTIONS BETWEEN SUBSTRATES AND THEIR between all pairwise comparisons using vegdist{vegan} function SURROUNDING WATER in R (Oksanen et al., 2012). We then calculated each habitat’s mul- Isolating individual biotic (hosts) and abiotic substrates along tidimensional “centroid” using the median value of each sample with their surrounding water allowed for direct comparisons within a habitat across all principal components. The distance of attached vs. free-living Vibrio communities. Fish hosts of each sample to its habitat centroid was calculated across all (FreshwaterFish and MarineFish) on average showed a nearly principal components. The variance around the median value 20-fold enrichment of total Vibrio relative abundance compared of sample-centroid distances was then compared across habitats to their surrounding water (30% ± SE 2.9 in fish vs. 1.6% ± in a standard ANOVA, followed by pairwise Tukey’s Honestly SE 0.37 in water, Table 1), but samples from sponges or plas- Significant Difference (HSD) tests. This entire process, from tic showed no significant enrichment [although a non-significant centroid calculation to HSD tests was implemented using the trend of enrichment was evident for Sponge habitats (Table 1)]. betadisper{vegan} function in R. To visualize our results, we plot- Despite this enrichment in fish hosts, we could not differen- ted each sample along their first two principal components, and tiate Vibrio community structure in fish microbiome samples plotted the multidimensional centroid. We then drew covariance and their surrounding environment with ANOSIM analyses, so ellipsoids around each habitat to illustrate the median distance long as comparisons were made within the same salinity category for all samples in a habitat around its centroid. Median sample- (Marine and Freshwater). centroid distances for each habitat were also plotted to better Interestingly, although Vibrio communities between fish visualize the within-habitat variance. and their surrounding water at a given salinity were statistically indistinguishable, communities between fresh and RESULTS marine salinity environments showed dramatic differences in OLIGOTYPE DISTRIBUTION ACROSS BIOTIC AND ABIOTIC both community structure and relative abundance of total SUBSTRATES Vibrio (Figure 2, Middle). Oligotypes 2, 4, and 8 dominated Oligotyping analysis of substrate associated-habitats (fish, both FreshwaterFish and FreshWater (cumulative abundance in sponges, plastics) yielded 74 unique oligotypes across 104 samples FreshwaterFish/FreshWater = 87.2%/71.9%), while Oligotypes 1, from 1,543,415 initial sequences. The minimum relative abun- 3, and 6 dominated MarineFish and MarineWater (cumulative dance threshold removed 808 rare oligotypes. The most abundant abundance in MarineFish/MarineWater = 60%/53% (Figure 3). 5 oligotypes represented 76% of the reads, with the top 10 repre- Experimental aquaria without fish at low salinity (water only) senting 94% (Table 2). Oligotypes that were abundant in at least showed a strong dominance of Oligotype 2, 4, and 8 (cumulative

Frontiers in Microbiology | Systems Microbiology November 2014 | Volume 5 | Article 563 | 6 Schmidt et al. Vibrio community structure across habitats

FIGURE 1 | Nonmetric Multidimensional Scaling (NMDS) plot of type. Pairwise ANOSIM permutation tests reveal all host habitats can be host-associated Vibrio communities based on oligotype distributions. signiﬁcantly differentiated except MarineFish and Sponge communities Labels are located at the center of covariance ellipsoids around each host (Table S1).

Table 4 | (A,B) SIMPER analysis output for the “All substrate” analysis grouping.

(A) Within habitat markers FreshwaterFish Cont (%) MarineFish Cont (%) Sponge Cont (%) Plastic Cont (%)

Oligo2 72.85 Oligo1 32.85 Oligo1 37.45 Oligo5 69.9 Oligo4 11.61 Oligo6 20.49 Oligo6 18.67 Oligo3 11.07 Oligo8 6.85 Oligo3 15.16 Oligo4 15.73 Oligo15 5.53 Oligo4 12.85 Oligo2 6.36 Oligo7 7.31 Oligo8 5.43 Oligo5 4.98 Oligo7 4.54 Oligo3 2.7 (B) Between habitat markers FreshwaterFish and MarineFish FreshwaterFish and Sponge FreshwaterFish and Plastic Average dissimilarity = 83.6 Average dissimilarity = 76.16 Average dissimilarity = 90.26

Oligo2 33.64 Oligo2 33.5 Oligo2 30.08 Oligo1 15.51 Oligo1 14.93 Oligo5 24.16 Oligo6 11.57 Oligo8 12.8 Oligo8 8.38 Oligo4 11.1 Oligo4 11.67 Oligo4 8.02 Oligo8 8.72 Oligo6 9.46 Oligo15 5.21 Cumulative 80.54 Cumulative 82.36 Cumulative 75.85

MarineFish and Sponge MarineFish and Plastic Sponge and Plastic Average dissimilarity = 63.21 Average dissimilarity = 85.02 Average dissimilarity = 89.48

Oligo1 19.77 Oligo5 23.41 Oligo5 23.33 Oligo6 16.14 Oligo1 16.07 Oligo1 13.31 Oligo4 15.01 Oligo6 12 Oligo6 8.46 Oligo8 8.77 Oligo4 9.34 Oligo4 8.18 Oligo2 8.45 Oligo7 7.04 Oligo2 7.64 Cumulative 68.14 Cumulative 67.86 Cumulative 60.92

A: The percent contribution of each oligotype to within-habitat Bray-Curtis similarity is shown (Cont%). B: The percent contribution of each oligotype to Bray-Curtis dissimilarities between two habitats is shown, along with average Bray-Curtis dissimilarities.

www.frontiersin.org November 2014 | Volume 5 | Article 563 | 7 Schmidt et al. Vibrio community structure across habitats

across all salinities, in water, fish and control samples, at greater than 10% relative abundance. Sponge associated Vibrio communities were also statistically indistinguishable from their surrounding water (ANOSIM P = 0.31), although this may have been in part due to the sample size difference between sponge and associated seawater samples (Sponge = 49, Associated seawater = 11), and the high vari- ability of particular oligotypes in some Sponge samples. Sponges showed significantly smaller relative abundance of Oligotype 7 and 5, and significantly increased abundance of Oligotypes 8 and 2(PairwiseT-tests P < 0.01 in all cases). Sponge associated Vibrio communities also showed no clear groupings according to species (Figure 2, Top). Lastly, Vibrio communities from plastic substrates overlapped completely with seawater communities collected alongside them (Figure 2, Bottom), and Oligotypes 3, 5, and 12, dominated both Plastics and associated seawater. We found no oligotype to be significantly enriched on plastic samples compared to their surrounding water, nor were there significant increases in total Vibrio on plastic substrates. In several cases, we found a single oligotype that did not occur in the surrounding water but dominated in relative abundance on an individual plastic substrate. This pattern was particularly apparent with Oligotypes 13, 2, and 7, which reached extremely high relative abundances on multiple occasions (e.g., Oligotype 2 at 91% relative abundance in the “10/09-Plastic” sample) (Figure 4).

OLIGOTYPE DISTRIBUTIONS ACROSS BROAD HABITAT TYPES In order to gain as broad a view as possible of Vibrio oligotype distributions across habitats, we included 179 samples from 7 environmental and host-associated habitats spanning a wide range of environmental and geographical gradients (Figure 5). This analysis yielded 99 oligotypes, of which the top 5 represented 65% of all reads, while the top 10 represented 90%. We observed the top 10 oligotypes from this analysis at high abundance in previous analyses (as determined by identical representative sequences), except Oligotype 10, which was novel to Sand-PAH mesocosms and is a highly divergent oligotype which could not be fully resolved. Sand-PAH samples were taken from beach sand communities near the Deepwater Horizon oil spill, and were likely enriched for PAH-associated species (Kappell et al., 2014). FIGURE 2 | Samples from the three host-associated studies included in All oligotypes that were highly abundant in a single sam- this meta-analysis separated out into individual NMDS plots alongside ple (>10% relative Vibrio abundance) occurred across all other water samples collected as part of the same study. Top: Sponges are habitat types. Abundant oligotypes were therefore also likely to labeled according to species. Water was collected next to Mycale sp. and Hexadella cf. dedritifera samples only. Middle: Fish water was sampled be common across a wide variety of habitat types (Figure 6). directly from aquaria in which fish were housed, in addition to “control” Oligotype 5 in particular was found to be both highly abundant water that was sampled from aquaria at an identical salinity, but without and frequent, occurring in all 179 samples analyzed across all any fish. Bottom: Ocean surface water was collected at the same time and habitats (Figure 6). Several oligotypes did not follow this gen- location as corresponding plastics. ANOSIM permutation tests show that eral pattern, and despite a relative ubiquity, they maintained at Fish, Sponge and Plastic associated Vibrio communities cannot be statistically separated out from their surrounding water environment. low mean relative abundances across all the samples in which Labels for each habitat are within their respective covariance ellipsoid. they occurred (e.g., Oligotypes 1, 2, 13, and 14, Figure 6). Comparing the mean relative abundance of individual oligotypes between marine hosts (Sponge, MarineFish) and marine environ- abundance = 67%), showing strong similarities to those aquaria ments (Seawater, Saltmarsh, Sand-PAH) revealed that 10 of the that did house fish. However, marine aquaria without fish showed top 17 oligotypes (Table 3) were significantly different between a strong dominance of Oligotype 2, inconsistent with marine these habitat categories at the bonferroni-corrected alpha level of aquaria that did contain fish. Interestingly, Oligotype 4 was found 0.0029. SIMPER revealed the strong influence of Oligotype 3 and

Frontiers in Microbiology | Systems Microbiology November 2014 | Volume 5 | Article 563 | 8 Schmidt et al. Vibrio community structure across habitats

FIGURE 3 | Oligotype distribution for FreshwaterFish and MarineFish bars (top, red-dashed lines are median values for each host-habitat). A clear tissue (FreshwaterFish: 0-x-i and 5-x-ii, MarineFish: 18-x-i and 18-x-ii), division between low and high salinity samples is seen, despite considerable associated water samples (x-x-W, blue lines), and water samples from variation within salinities. Significant differences in median total Vibrio relative aquaria containing no fish (x-C-W). The relative abundance of each abundance exists between FreshwaterFish and MarineFish samples, and oligotype within the total Vibrio diversity for each sample is shown in stacked between MarineFish samples and their surrounding water. The Vibrio bar graphs (bottom), and the proportion of the total Vibrio (relative community of fish food used during experimental period is also shown abundance) within all bacterial diversity for each sample is shown with gray (“FOODX”).

FIGURE 4 | Oligotype distributions for Plastic samples and their surface. Black arrows indicate plastic samples that contain a single oligotype surrounding water. Sample labels indicate the date the samples were at greater than 50% relative abundance. “OligotypeNA” represents an collected and the sample type. Water and plastic samples collected on the oligotype that was not among the top 10 most abundant oligotypes from any same date are associated with one another. All water was collected at the three of our oligotype groupings.

5 in Saltmarsh and non-host associated Seawater communities, Analysis of “within-habitat similarity” (a measure of sam- and Oligotype 5 in Sand-PAH mesocosms. Oligotype 5 was often ple dispersion within a habitat) showed significant differences extremely abundant in seawater samples, including those from between habitats. Post-hoc Tukeys pairwise tests revealed Sand- a Vibrio bloom (from project ICM_PML_Bv6, Table 1), and on PAH mesocosms (median distance to centroid = 0.047) and plastic samples (Figure 4). ANOSIM analyses revealed that across Saltmarsh (median distance to centroid = 0.046) contained sig- all habitat pairwise comparisons only Sponge and MarineFish, nificantly less variance than Seawater (median distance to cen- Seawater and Saltmarsh, Sand-PAH and Plastic, and Sand-PAH troid = 0.388) and Sponge (median distance to centroid = 0.49) and Seawater could not be significantly differentiated from one habitats. We note here however, that these results do not control another (Table S1). for the larger geographic area over which Seawater and Sponge

www.frontiersin.org November 2014 | Volume 5 | Article 563 | 9 Schmidt et al. Vibrio community structure across habitats

MEGABLAST queries of abundant oligotypes returned on average 120 perfect matches from the NCBI nr database, although variance in this number was high (ranging from 0 to 1000). No oligotype with more than two perfect NCBI hits came from only one source, and any oligotype with more than 50 perfect NCBI hits was isolated from at least three different sources. Oligotype 5, which was the most abundant across our habitats (Figure 6), also had the maximum number of allowable perfect hits (1000). Which habitat a query sequence came from was an extremely poor predictor for the isolation source of its NCBI hits, although Oligotypes 4 and 8, which contributed to FreshwaterFish had previously been isolated from freshwater environments (Table 3), and both matched V. cholerae, found in brackish water, while abundant oligotypes in marine organisms (e.g., 1 and 6), did FIGURE 5 | NMDS plot with covariance ellipsoids for both not return any previous isolations from freshwater environments. host-associated and environmental samples. Sample names refer to habitat type and VAMPS project listed in Table 1. Surprisingly, Oligotype 2, which contributed most to freshwater environments, had no perfect hits from freshwater sources. The majority of isolation sources from all our NCBI hits (76%) were marine hosts, although we note the potential for database biases.

DISCUSSION Our results represent the first attempt to use subtle nucleotide variation at a single, 60 bp gene marker to make sense of community level patterns across habitats within the genus Vibrio. Although our results are not the first to explore Vibrio community patterns across diverse habitats (Hunt et al., 2008; Preheim et al., 2011a; Szabo et al., 2013), we use preexisting sequence data to extend conclusions made by previous authors to a broader survey of unexplored habitats and provide novel insights into substrate-associated communities and their surrounding water. The breadth and scope of our analyses, 211 samples representing seven unique habitats, revealed or supported several interesting patterns suggestive of two broad hypotheses regarding different aspects of Vibrio ecology and life history. First, Vibrio FIGURE 6 | Commonness and abundance plot of all oligotypes that were part of the mixed habitat sample grouping analysis (Table 2). The contains many generalist taxa, each adapted to a wide range of occurrence (presence/absence) of each of the 99 oligotypes across all 179 animal hosts. Second, our data suggest even these “host-adapted” samples is plotted along the x-axis while its mean relative abundance across vibrios occur as members of free-living communities facilitat- all samples is plotted on the y-axis. Samples that are both common and ing long distance dispersal to disparate hosts. We suggest both abundant are found in the top right, while those that are common, but rare are in the bottom right. Both rare and uncommon are found in the bottom of these characteristics, combined with previous understand- left. The top 17 most abundant oligotypes from Table 3 are also labeled. ingofrapidgrowthrates(McDonough et al., 2013; Skorupski and Taylor, 2013) fit the description of an “r-strategist” life history. samples were collected as compared to Sand-PAH and Saltmarsh Several patterns within our data support these suggestions. samples. All other pairwise comparisons were insignificant at the Fish acclimated to marine salinities share highly similar Vibrio 0.05 alpha level after multiple comparison adjustments. communities with geographically and phylogenetically distinct sponges (Figure 1, Table 4). Both marine acclimated fish and PHYLOGENETIC AND METADATA ANALYSIS OF ABUNDANT sponge communities were typified by a strong dominance of OLIGOTYPES Oligotypes 1, 4, and 6, all of which were above 15% mean relative Our phylogenetic analysis of ARB reference sequences revealed no abundance in both “Sponge” and “MarineFish” samples. These 16S rRNA gene phylogeny-habitat relationship. This is evidenced oligotypes contributed at least 12% of within group Bray-Curtis by the spread of reference sequences from the same isolation similarities for each habitat (Table 4). Furthermore, ANOSIM source across the phylogenetic tree (Figure 7). Our 17 abundant analysis at the community level (i.e., using all oligotypes in each oligotypes were also not monophyletic according to the habi- community) found these habitats could not be significantly sep- tat in which they were most abundant (e.g., sponges or fish). arated, a conclusion supported by their overlapping distribution For example, oligotype sequences dominant in FreshwaterFish in 2D projections (Figures 1, 5). Although these communities do and Freshwater (Oligotype 2, 4, and 8) were found to branch in not appear to be host specific, they do appear to be specific to different parts of our phylogenetic tree (Figure 7). biotic hosts. Plastic communities, on an abiotic substrate, were

Frontiers in Microbiology | Systems Microbiology November 2014 | Volume 5 | Article 563 | 10 Schmidt et al. Vibrio community structure across habitats

FIGURE 7 | PhyML phylogeny of full-length 16S rRNA gene sequences strain is color coded at its terminal node. The proportion of MEGABLAST from type strain isolates in the SILVA ARB database, with the 17 hits that fell into each isolation source category is coded for the 17 most abundant oligotypes found in this study added by Maximum oligotype sequences added to the tree, and shown in bar graphs at each Parsimony over the 60 bp region. The isolation source for each type oligotype node. typifiedbyOligotypes5,3,and12,andwhileoligotypesthat on Plastic. This suggests that vibrios behave differently with typified marine fish and sponges sometimes occurred on plas- respect to adaptation to and colonization of biotic vs. abiotic tics, they were always in low abundance. Furthermore, pairwise substrates. Furthermore, the finding that Vibrio oligotypes asso- ANOSIM tests showed that although Sponges and MarineFish ciated with plastics overlap with those in the surrounding sea- could not be significantly separated both could be separated from water suggests that recruitment may take place far from the Plastic. We confirmed this result with significant ANOSIM group- origins of the plastic marine debris itself, typically thought to be ings of MarineFish and Sponge datasets pooled together, at the land-based. exclusion of Plastic samples (data not shown). We found group- Explaining the similarity between Vibrio communities in fish ings were again characterized by high abundance of Oligotypes and sponges is challenging, but overlap in habitat geography 1, 4, and 6 on biotic substrates, and Oligotype 5, 3, and 12 during sampling, or overlap of laboratory sample preparation

www.frontiersin.org November 2014 | Volume 5 | Article 563 | 11 Schmidt et al. Vibrio community structure across habitats

in space or time can be ruled out. Sponges were collected by average relative abundance in environmental samples (Seawater, SCUBA and Remote Operated Vehicle (ROV) deep-sea dives Saltmarsh) and an abiotic substrate (Plastic) of >40%, while from the Northeast Atlantic and Mediterranean Sea (Reveillaud averaging only 2.8% in hosts. This oligotype was found across et al., 2014) from 1981 to 2011, while all fish were col- all 179 samples analyzed as part of this study (Figure 6), and lected from experimental aquaria filled with sterilized water and was widely represented in our MEGABLAST results (Table 3). Instant Ocean® salt mix in a Woods Hole, MA laboratory in Analysis of Vibrio sequences recovered from an earlier study of 2013 (Schmidt et al., submitted; Supplementary Data Sheet 1). plastic marine debris samples (Zettler et al., 2013) also detected Furthermore, although sequencing was conducted on the same Oligotype 5 (data not shown) despite employing a different Illumina HiSeq instrument, run dates were several months apart, sequencing platform. Together, these data suggest a widely dis- and sample storage occurred in different freezers, making contributed, predominantly “non-host associated” Vibrio found in tamination between projects unlikely. In addition to community hosts only through chance or ephemeral colonization. level patterns outlined above, our MEGABLAST analysis of the representative sequences from oligotypes most representative of SALINITY DRIVES VIBRIO STRUCTURE IN WATER AND HOST our sponge and fish samples revealed that each was previously COMMUNITIES isolated from a variety of marine hosts including crabs, jelly- Salinity is a known driver of Vibrio community structure and fish, sea squirts, corals, fish, clams, and sea cucumbers (Table 3, most vibrios are thought to occur in brackish or marine envi- Table S2). Lastly, a previous comparison of two phylogenetically ronments (Takemura et al., 2014). Schmidt et al. (submitted) distinct hosts (mussels and crabs) also found significant over- (Supplementary Data Sheet 1) experimentally manipulated salt lap in Vibrio communities (Preheim et al., 2011a), although we concentrations in aquaria containing a euryhaline (salt tolerant) note in this case hosts were not geographically or temporally fish, and characterized the resulting bacterial community. They distinct. found that communities in both the fish and water changed Such a large degree of host plasticity does not appear to be across the salt gradient, but that they did not change concur- a ubiquitous feature among all bacterial taxa, and Vibrio clearly rently, resulting in drastically different communities in fish and differs from other taxa at the community level. A previous study tank water. Interestingly, fish/water differences at high salinities on sponge-microbe associations demonstrated host-specificity (18 and 30 ppt) were in part driven by high Vibrio relative abun- within the genus Nitrospira (Reveillaud et al., 2014). This study dance in fish, vs. its relative rarity in tank water (Figure 3). Vibrio used the same oligotyping pipeline of V6 sequences, from the also partly drove differences in fish microbiomes across the salin- same samples analyzed here, and the authors showed strong host ity gradient, with a nearly 10-fold increase in total Vibrio relative specificity of Nitrospira oligotypes to sponges at the species level abundance from 0 to 30 ppt acclimated fish. This study did not, (see Figure 4 in Reveillaud et al., 2014). They found closely related however, resolve bacteria below the genus level, and could not sponge species had differential enrichment preferences for closely make conclusions about variation within a genus across salini- related Nitrospira phylogenetic lineages across varying bathymet- ties. The study therefore did not assess if increases in total Vibrio ric and geographic areas. Our oligotyping analysis, focusing on relative abundance up the gradient were due to the same taxa Vibrio, highlighted the lack of sponge-specific patterns within this becoming more abundant, or to the addition of novel taxa at genus (Figure 2, Top), and is therefore in stark contrast to the higher salinities. Nor were they able to assess if Vibrio inside fish patterns illustrated for Nitrospira. were the same as those found in the tank water. Further supporting our suggestion that taxa within the genus The fine scale resolution provided by our oligotyping anal- Vibrio are generalist, long distance dispersing, r-strategists is yses allowed us to answer these questions, and we show that the commonality of some oligotypes across the broad range of Vibrio community structure between water and fish are broadly host associated and environmental habitat types in this study consistent, with both habitats sharing similar occurrence and (Figure 7). Although Vibrio does not form spores (Madigan et al., relative abundances of particular Vibrio oligotypes (Figure 3), 2009), it is known to enter a “viable but uncultivable” state despite an overall enrichment of Vibrio relative abundance in fish under stressful or nutrient limiting conditions (Ramaiah et al., vs. water. We also show that FreshWater (tank water commu- 2002). Research with the squid symbiont Vibrio fischeri found nity) and FreshwaterFish (fish microbiome community) cannot the bacteria quickly became uncultivable and non-luminescent be significantly distinguished with ANOSIM tests, and SIMPER in nutrient poor water outside of its host, but retained the abil- analyses find the same oligotypes (2, 4, and 8) are represen- ity to colonize, and luminesce, given re-entry to a suitable host tative of both FreshWater and FreshwaterFish (Table 4). The (Lee and Ruby, 1995). Our results demonstrate that Oligotypes same is true for comparisons between high salt acclimated fish 1, 6, 4, 8, and 9 were all significantly enriched in marine hosts (MarineFish) and their water (MarineWater), which are both compared to marine environmental samples, with as much as characterizedbyOligotypes1,6,3,and4.Thistrendforboth a 168-fold enrichment, yet they all occurred at low abundance salinities is evident from 2D projections of community structure in open-ocean, host-independent samples. It is possible the rare (Figure 2, Middle), which show overlapping ellipsoids of fish and occurrence of these “host-associated” oligotypes in seawater sam- water habitats (although some separation is evident). Despite an ples represent taxa that have entered viable but uncultivable states, overall shift in community structure across the salinity gradi- giving them the ability to disperse long distances in nutrient ent, Oligotype 4 remains at high abundance in both fresh and poor waters between opportunistic colonization of a wide vari- marine samples. Oligotype 4 was highly enriched in all host- ety of marine hosts. Conversely, Oligotype 5 was found at an associated samples (Sponges, FreshwaterFish, MarineFish), and

Frontiers in Microbiology | Systems Microbiology November 2014 | Volume 5 | Article 563 | 12 Schmidt et al. Vibrio community structure across habitats

extremely rare in environmental water samples not associated short rRNA gene sequence, our analyses show strong convergence with a host collection (i.e., completely independent of hosts). between host-associated communities, despite wide geographic Furthermore, MEGABLAST results from this oligotype show and phylogenetic distance between them. We also show a surpris- its previous isolation from both marine and freshwater hosts, ing overlap, and a lack of significant divisions, between Vibrio but never from seawater (Table 4). Together these results are communities in hosts and those found in their surrounding suggestive of host-associated, potentially salt-tolerant Vibrio taxa. aquatic environments. Our results further support that Vibrio, as a genus, is largely populated by generalist r-strategist species, ADVANTAGES AND LIMITATIONS OF OLIGOTYPING ANALYSIS FOR capable of long distance dispersal, a wide breadth of growth VIBRIO requirements, and rapid growth rates (Szabo et al., 2013). By separating regions of a marker gene that contain biologically meaningful variation from stochastic error, oligotyping allows ACKNOWLEDGMENTS single nucleotide differences across short marker genes to iden- We would like to thank the VAMPS bioinformatics team, princi- tify potentially ecologically meaningful patterns with extremely pally Andy Voorhis and Anna Shipunova. We also thank Joseph small amounts of information (Eren et al., 2013a). This tech- Migliore with the Rhode Island Department of Environmental nique provides substantial benefits, avoiding the need for lengthy Management for assistance collecting samples from Mt. Hope and expensive culturing and sequencing protocols, and allows Bay. Open ocean plastic samples were collected by students researchers to tap into massive existing databases to ask novel eco- and staff at the SEA Education Association/SEA Semester in logical questions at high-throughput levels across global scales. Woods Hole. This work was supported by an NSF Collaborative We note, however, that some serious limitations do exist for these grant to Erik Zettler (OCE-1155379), Tracy J. Mincer (OCE- type of data. 16S rRNA gene sequences can be nearly identical 1155671) and Linda A. Amaral-Zettler (OCE-1155571), NSF across multiple Vibrio species (Gomez-Gil et al., 2004), and even TUES grant to Erik Zettler and Linda A. Amaral-Zettler (DUE- contain variance between copies of 16S rRNA genes within a sin- 1043468). Additional support came from the Woods Hole Center gle genome, making its use as a phylogenetic tool difficult or for Oceans and Human Health from the National Institutes of impossible. In addition, because our analyses use only 60 bp of Health and National Science Foundation (NIH/NIEHS 1 P50 DNA sequence at a single marker gene, our data are insufficient ES012742-01 and NSF/OCE 0430724-J: Linda A. Amaral-Zettler for any phylogenetic inference, and we cannot deduce relatedness and Leslie Murphy) and an NSF/OCE-1128039 award (Linda A. between individual oligotypes. This provides major limitations in Amaral-Zettler and Leslie Murphy). Victor Schmidt was sup- our ability to address some hypotheses about the evolutionary ported during this work by an NSF IGERT fellowship (DGE history of vibrios adapted to specific habitats. We could not inves- 0966060, Dr. David Rand, PI). tigate, for example, if the abundant oligotypes of fish and sponges (Oligotypes 1, 4, and 6) share common ancestry, which would SUPPLEMENTARY MATERIAL suggest speciation within the genus after association with the The Supplementary Material for this article can be found host. We also cannot confidently tie our conclusions to previous online at: http://www.frontiersin.org/journal/10.3389/fmicb. observations about particular Vibrio species, such as V. splen- 2014.00563/abstract didus’ potentially recent adaption to particulate adhesion (Hunt et al., 2008), or V. cholerae’s affinity for freshwater (Skorupski REFERENCES Aiso, K., Simidu, U., and Hasuo, K. (1968). Microflora in the digestive tract of and Taylor, 2013), since we cannot make taxonomic assignments inshore fish in Japan. Microbiology 52, 361–364. to any of our oligotypes. High-resolution taxonomic assignment Andrews, J., and Harris, R. (1986). “r- and K-selection and microbial ecology,” in of Vibrio has been a significant challenge, necessitating the use Advances in Microbial Ecology,edK.C.Marshall(NewYork,NY:Springer), of genomic analyses including DNA-DNA hybridization, multi- 99–147. locus sequence analysis (MLSA), and genome sequencing for Austin, B., and Austin, D. (2007). Bacterial Fish Pathogens; Diseases of Farmed and Wild Fish, 4th Edn. Chichester: Praxis Publishing. species- or strain-level identification (Thompson et al., 2005). Chin, C., Sorenson, J., Harris, J. B., Robins, W. P., Charles, R. C., Jean-Charles, R. Analyses of multiple loci, or entire genome sequences, are there- R., et al. (2011). The origin of the Haitian cholera outbreak strain. N. Engl. J. fore required to make any phylogenetic inference (Thompson Med. 364, 33–42. doi: 10.1056/NEJMoa1012928 et al., 2005; Preheim et al., 2011b). However, this study shows the Cordero, O. X., Wildschutte, H., Kirkup, B., Proehl, S., Ngo, L., Hussain, F., et al. utility of oligotyping as an easily adaptable, high-throughput and (2012). Ecological populations of bacteria act as socially cohesive units of antibiotic production and resistance. Science 337, 1228–1231. doi: 10.1126/sci- unbiased method for large-scale analyses of data from publically ence.1219385 available sequence data repositories, and suggests its wide appli- Eren, M., Maignien, L., Sul, W. J., Murphy, L. G., Grim, S. L., Morrison, H. G., cation could greatly extend the range of possibilities to explore et al. (2013a). Oligotyping: differentiating between closely related microbial taxa microbial ecology studies of particular genera. using 16S rRNA gene data. Methods Ecol. Evol. 4, 1111–1119. doi: 10.1111/2041- 210X.12114 Eren, M., Sogin, M. L., Morrison, H. G., Vineis, J. H., Fisher, J. C., Newton, R. J., CONCLUSIONS et al. (2014). A single genus in the gut microbiome reflects host preference and Our analysis combines a novel bioinformatics technique with specificity. ISME J. doi: 10.1038/ismej.2014.97. [Epub ahead of print]. large quantities of Vibrio 16S rRNA gene sequence data to reveal Eren, M., Vineis, J. H., Morrison, H. G., and Sogin, M. L. (2013b). A filtering method to generate high quality short reads using Illumina paired-end patterns of Vibrio ecology across a wide range of environmental, technology. PLoS ONE 8:e66643. doi: 10.1371/journal.pone.0066643 host, and abiotic substrate-associated habitats. Despite the draw- Eren, M., Zozaya, M., Taylor, C. M., Dowd, S. E., Martin, D. H., and Ferris, M. backs for phylogenetic and taxonomic inference of using a single, J. (2011). Exploring the Diversity of Gardnerella vaginalis in the genitourinary

www.frontiersin.org November 2014 | Volume 5 | Article 563 | 13 Schmidt et al. Vibrio community structure across habitats

tract microbiota of monogamous couples through subtle nucleotide variation. marine invertebrates. Environ. Microbiol. 13, 265–275. doi: 10.1111/j.1462- PLoS ONE 6:e26732. doi: 10.1371/journal.pone.0026732 2920.2010.02328.x FAO. (2012). The State of World Fisheries and Aquaculture. Rome: Food and Preheim, S. P., Timberlake, S., and Polz, M. F. (2011b). Merging taxonomy with Agriculture Organization of the United Nations. ecological population prediction in a case study of Vibrionaceae. Appl. Environ. Gilbert, J. A., Steele, J. A., Caporaso, J. G., Steinbrück, L., Reeder, J., Temperton, B., Microbiol. 77, 7195–7206. doi: 10.1128/AEM.00665-11 et al. (2012). Defining seasonal marine microbial community dynamics. ISME Ramaiah, N., Ravel, J., Straube, W. L., Hill, R. T., and Colwell, R. R. (2002). J. 6, 298–308. doi: 10.1038/ismej.2011.107 Entry of Vibrio harveyi and Vibrio fischeri into the viable but noncultur- Gomez-Gil, B., Soto-Rodríguez, S., García-Gasca, A., Roque, A., Vazquez-Juarez, able state. J. Appl. Microbiol. 93, 108–116. doi: 10.1046/j.1365-2672.2002. R., Thompson, F. L., et al. (2004). Molecular identification of Vibrio harveyi- 01666.x related isolates associated with diseased aquatic organisms. Microbiology 150, Reveillaud, J., Maignien, L., Murat, E., A., Huber, J., Apprill, A., Sogin, 1769–1777. doi: 10.1099/mic.0.26797-0 M. L., et al. (2014). Host-specificity among abundant and rare taxa Heidelberg, J. F., Eisen, J., Nelson, W. C., Clayton, R., Gwinn, M. L., Dodson, R. in the sponge microbiome. ISME J. 8, 1198–1209. doi: 10.1038/ismej. J., et al. (2000). DNA sequence of both chromosomes of the cholera pathogen 2013.227 Vibrio cholerae. Nature 406, 477–483. doi: 10.1038/35020000 Ruby, E., and Nealson, K. (1976). Symbiotic association of Photobacterium fischeri Hunt, D. E., David, L. A., Gevers, D., Preheim, S. P., Alm, E. J., and Polz, with the marine luminous fish Monocentris japonica:amodelofsymbiosisbased M. F. (2008). Resource partitioning and sympatric differentiation among on bacterial studies. Biol. Bull. 151, 547–586. closely related bacterioplankton. Science 320, 1081–1085. doi: 10.1126/science. Skorupski, J., and Taylor, R. (2013). “Toxin and virulence regulation in Vibrio 1157890 cholerae,” in Regulation of Bacterial Virulence, eds M. Vasil and A. Darwin Huse, S. M., Dethlefsen, L., Huber, J., Mark Welch, D., Welch, D. M., Relman, (Washington, DC: ASM Press), 241–262. D., et al. (2008). Exploring microbial diversity and taxonomy using SSU Szabo, G., Preheim, S. P., Kauffman, K. M., David, L., Shapiro, J., Alm, E. J., rRNA hypervariable tag sequencing. PLoS Genet. 4:e1000255. doi: 10.1371/jour- et al. (2013). Reproducibility of Vibrionaceae population structure in coastal nal.pgen.1000255 bacterioplankton. ISME J. 7, 509–519. doi: 10.1038/ismej.2012.134 Huse, S. M., Mark Welch, D. B., Voorhis, A., Shipunova, A., Morrison, H. G., Eren, Takemura, A. F., Chien, D. M., and Polz, M. F. (2014). Associations and dynamics of A. M., et al. (2014). VAMPS: a website for visualization and analysis of microbial Vibrionaceae in the environment, from the genus to the population level. Front. population structures. BMC Bioinformatics 15:41. doi: 10.1186/1471-2105-15- Microbiol. 5:38. doi: 10.3389/fmicb.2014.00038 41 Thompson, F. L., Gevers, D., Thompson, C. C., Dawyndt, P., Hoste, B., Munn, Kappell, A. D., Wei, Y., Newton, R. J., Van Nostrand, J. D., Zhou, J., McLellan, S. L., C. B., et al. (2005). Phylogeny and molecular identification of Vibrios on the et al. (2014). The polycyclic aromatic hydrocarbon degradation potential of Gulf basis of multilocus sequence analysis phylogeny. Appl. Environ. Microbiol. 71, of Mexico native coastal microbial communities after the Deepwater Horizon oil 5107–5115. doi: 10.1128/AEM.71.9.5107 spill. Front. Microbiol. 5:205. doi: 10.3389/fmicb.2014.00205 Zettler, E. R., Mincer, T. J., and Amaral-Zettler, L. A. (2013). Life in the Lee, K., and Ruby, E. G. (1995). Symbiotic role of the viable but nonculturable “Plastisphere”: microbial communities on plastic marine debris. Environ. Sci. state of Vibrio fischeri in Hawaiian coastal seawater. Appl. Environ. Microbiol. 61, Technol. 47, 7137–7146. doi: 10.1021/es401288x 278–283. Letunic, I., and Bork, P. (2011). Interactive Tree Of Life v2: online annotation and Conflict of Interest Statement: The authors declare that the research was con- display of phylogenetic trees made easy. Nucleic Acids Res. 39, W475–W478. doi: ducted in the absence of any commercial or financial relationships that could be 10.1093/nar/gkr201 construed as a potential conflict of interest. Liston, J. (1956). Quantitative variations in the bacterial flora of flatfish. J. Gen. Microbiol. 15, 305–314. Received: 01 August 2014; accepted: 07 October 2014; published online: 13 November Madigan, M., Martinko, J., Dunlap, P., and Clark, D. (2009). Brock Biology of 2014. Microorganisms, 12th Edn. San Francisco, CA: Pearson Benjamen Cummings. Citation: Schmidt VT, Reveillaud J, Zettler E, Mincer TJ, Murphy L and Amaral- McDonough, E., Bradley, E., and Camilli, A. (2013). “Regulating the transition of Zettler LA (2014) Oligotyping reveals community level habitat selection within the Vibrio cholerae out of the host,” in Regulation of Bacterial Virulence, eds E. Vasil genus Vibrio. Front. Microbiol. 5:563. doi: 10.3389/fmicb.2014.00563 and A. Darwin (Washington, DC: ASM Press), 566–578. This article was submitted to Systems Microbiology, a section of the journal Frontiers Mutreja, A., Kim, D. W., Thomson, N. R., Connor, T. R., Lee, J. H., Kariuki, S., et al. in Microbiology. (2011). Evidence for several waves of global transmission in the seventh cholera Copyright © 2014 Schmidt, Reveillaud, Zettler, Mincer, Murphy and Amaral-Zettler. pandemic. Nature 477, 462–465. doi: 10.1038/nature10392 This is an open-access article distributed under the terms of the Creative Commons Oksanen, J., Blanchet, G., Kindt, R., Legendre, P., O’Hara, R., Simpson, G., et al. Attribution License (CC BY). The use, distribution or reproduction in other forums is (2012). Vegan: Community Ecology Package. R package version 1.17-3. permitted, provided the original author(s) or licensor are credited and that the original Preheim, S. P., Boucher, Y., Wildschutte, H., David, L., Veneziano, D., Alm, E. publication in this journal is cited, in accordance with accepted academic practice. No J., et al. (2011a). Metapopulation structure of Vibrionaceae among coastal use, distribution or reproduction is permitted which does not comply with these terms.

Frontiers in Microbiology | Systems Microbiology November 2014 | Volume 5 | Article 563 | 14